feat(voice): add streaming asr sidecar for v2 duplex

deployment/docker-compose.parallel.server.yml

@@ -22,6 +22,8 @@ x-app-env: &app_env
   KB_SERVICE_URL: http://kb-service:8000
   REPORTING_SERVICE_URL: http://reporting-service:8000
   SUPERVISOR_SERVICE_URL: http://supervisor-service:8000
+  AI_VOICE_V2_STREAMING_ASR_BASE_URL: http://streaming-asr-sidecar:8021
+  AI_VOICE_V2_STREAMING_ASR_TIMEOUT_SECONDS: "4"
 
 x-service-defaults: &service_defaults
   build:
@@ -123,8 +125,38 @@ services:
       <<: *app_env
       APP_MODULE: services.ai_orchestrator_service.app:app
 
+  streaming-asr-sidecar:
+    <<: *service_defaults
+    command: ["sh", "-c", "uvicorn ${APP_MODULE} --host 0.0.0.0 --port 8021"]
+    environment:
+      <<: *app_env
+      APP_MODULE: services.streaming_asr_sidecar_service.app:app
+      AI_VOICE_V2_STREAMING_ASR_MODEL: base
+      AI_VOICE_V2_STREAMING_ASR_COMPUTE_TYPE: int8
+      AI_VOICE_V2_STREAMING_ASR_DEVICE: cpu
+      AI_VOICE_V2_STREAMING_ASR_CACHE_DIR: /models/faster_whisper
+      AI_VOICE_V2_STREAMING_ASR_SUPPORTED_LANGUAGES: ru
+    volumes:
+      - ../.data_pg_parallel:/app/.data_pg_parallel
+      - ../.models/faster_whisper:/models/faster_whisper
+    healthcheck:
+      test:
+        - CMD
+        - python
+        - -c
+        - |
+          from urllib.request import urlopen
+          import sys
+          sys.exit(0 if urlopen("http://127.0.0.1:8021/health", timeout=2).getcode() == 200 else 1)
+      interval: 15s
+      timeout: 5s
+      retries: 10
+      start_period: 10s
+
   ai-voice-runtime-service:
     <<: *service_defaults
+    depends_on:
+      - streaming-asr-sidecar
     environment:
       <<: *app_env
       APP_MODULE: services.ai_voice_runtime_service.app:app